Days 1 to 35, from the NOAA GEFS extended ensemble, every day
GEFS subseasonal outlook
Anomalies against GEFS's own 2000–2019 reforecast at the same lead and time of year, so the model's drift is removed rather than read as a forecast. Weekly maps, the change since the run a week earlier and the polar vortex — one figure at a time, picked from the menu. The same layout as the GEPS subseasonal outlook, for comparing the two models.
← → step through the last option row · ↑ ↓ change product · click a map to enlarge · the address bar carries the selection, so a view can be linked.
How it is built
Each day's 00Z GEFS run is read member by member, once it reaches day 35, from NOAA's open-data bucket. Daily values are the mean of the 12Z and 00Z fields ending each day (precipitation summed and outgoing longwave averaged over the four six-hour windows), box-averaged to 1.5° and meaned by week. Exactly the same arithmetic is applied to the GEFSv12 reforecast — the 35-day Wednesday runs of 2000–2019, four members each, pooled within ±15 days of the start date — and each week of the forecast is differenced from the reforecast's mean at the same week of lead. That removes the model's drift and its mean bias together, which a comparison with a reanalysis would not. The anomalies are therefore against GEFS's 2000–2019 climate: 500 hPa height has its global mean removed each week so the circulation shows rather than two decades of warming; temperature and precipitation keep theirs. The change maps compare with the run exactly seven days earlier, whose weeks 2–5 cover the same dates as this run's weeks 1–4. The polar vortex is corrected the same way against MERRA-2.